Discovery of a novel dehydratase of the fatty acid synthase type II critical for ketomycolic acid biosynthesis and virulence of Mycobacterium tuberculosis

The fatty acid synthase type II (FAS-II) multienzyme system builds the main chain of mycolic acids (MAs), important lipid pathogenicity factors of Mycobacterium tuberculosis (Mtb). Due to their original structure, the identification of the (3 R)-hydroxyacyl-ACP dehydratases, HadAB and HadBC, of Mtb FAS-II complex required in-depth work. Here, we report the discovery of a third dehydratase protein, HadDMtb (Rv0504c), whose gene is non-essential and sits upstream of cmaA2 encoding a cyclopropane synthase dedicated to keto- and methoxy-MAs. HadDMtb deletion triggered a marked change in Mtb keto-MA content and size distribution, deeply impacting the production of full-size molecules. Furthermore, abnormal MAs, likely generated from 3-hydroxylated intermediates, accumulated. These data strongly suggest that HadDMtb catalyzes the 3-hydroxyacyl dehydratation step of late FAS-II elongation cycles during keto-MA biosynthesis. Phenotyping of Mtb hadD deletion mutant revealed the influence of HadDMtb on the planktonic growth, colony morphology and biofilm structuration, as well as on low temperature tolerance. Importantly, HadDMtb has a strong impact on Mtb virulence in the mouse model of infection. The effects of the lack of HadDMtb observed both in vitro and in vivo designate this protein as a bona fide target for the development of novel anti-TB intervention strategies.


M. tuberculosis holds a putative HadD ortholog that is not essential for survival. Protein-protein
BLAST searches performed against Mtb H37Rv genome 17 , using the MSMEG_0948 (HadD Msm ) protein sequence as a probe, showed the presence of a potential ortholog of HadD Msm with a sequence identity rate of 68% in Mtb, Rv0504c, which we named HadD Mtb (Fig. 1A). The latter, annotated as "conserved protein" and having a theoretical monomeric mass of 18.4 kDa 17 , bears similarly to HadD Msm a degenerate hydratase 2 motif 'F-x(2)-ax(2)-D-x(2)-P-x-H-x(5)-A' containing the putative catalytic dyad Asp (D37) and His (H42) (Fig. 1A). The chromosomal region of hadD gene is partially conserved between Mtb and M. smegmatis (Fig. 1B). Interestingly, cmaA2 (Rv0503c) gene sits downstream of hadD Mtb (Rv0504c) on Mtb chromosome and is transcribed in the same direction. It encodes the mycolic acid methyltransferase (MA-MT) CmaA2 that has a function of cyclopropane synthase and introduces a trans or cis cyclopropane at the proximal position of both keto-and methoxy-MAs 18 . The lack of a cmaA2 ortholog in M. smegmatis (Fig. 1B) is in agreement with the absence of these MA classes in this species 2 .
The essentiality of hadD Mtb gene was examined by generating an in-frame unmarked deletion (Fig. 1B) using a two-step homologous recombination method 19 , so that it does not cause any polar effect on cmaA2 expression. The gene deletion was verified by PCR analysis (Fig. 1C). The viability of the resulting mutant strain showed that hadD Mtb is not essential for the survival of Mtb in axenic culture, similar to the situation seen for hadD Msm in M. smegmatis 16 . This is consistent with the prediction of non-essentiality of hadD Mtb made from a microarray-based study 20 , but in discrepancy with a more recent survey using high-resolution phenotypic profiling where hadD Mtb was predicted to be essential 21 . HadD Mtb deletion alters Mtb physiology and virulence. In the aim of evaluating the importance of HadD Mtb function in the physiology of Mtb, phenotypic assays were realized. Although Mtb ΔhadD is viable, this mutant strain showed difficulties to grow in planktonic culture as compared to the wild type (wt) strain ( Fig. 2A). HadD Mtb inactivation also resulted in a strong alteration of the structuration of biofilms at the air-liquid interface, which appeared thinner with large clumps (Fig. 2B). There was a deep change of the colony morphology as well, with much larger, flat and spread colonies (Fig. 2C). The wt phenotype was restored in the mutant upon complementation meaning that these phenomena were linked to hadD Mtb deletion. The lack of HadD Mtb also conferred a high sensitivity to low temperature (Fig. 2D). In addition, Mtb ΔhadD was slightly more susceptible than Mtb wt to rifampicin, a first line antituberculous drug targeting the DNA-dependent RNA synthesis. In contrast, there was no significant difference in the sensitivity of both strains to isoniazid and ethambutol, two other firstline TB drugs, and to ciprofloxacin, a broad spectrum fluoroquinolone, as well as to the SDS detergent (Supplementary  Table S1).
To assess the influence of hadD Mtb on the virulence level of Mtb, infection trials were performed with severe combined immunodeficiency (SCID) mice. This murine model provides a rapid and sensitive method for evaluating in vivo growth characteristics of mycobacterial strains during the acute phase of infection [22][23][24][25] . Although overexpressing hadD Mtb gene had no significant impact on the virulence level, we observed marked and reproducible www.nature.com/scientificreports www.nature.com/scientificreports/ In conclusion, HadD Mtb plays a key role in the faculty of the tubercle bacillus to invade and multiply within the infected host. Furthermore, it is also important for the fitness and the capacities of Mtb bacilli to assemble into colonies or biofilms, and for their tolerance to low temperature. The deep changes observed in hadD Mtb -deficient strain likely are the consequences of an alteration of the cell envelope composition and architecture.
HadD Mtb influences the mycolic acid profile of Mtb. Mtb produces three main MA classes, α-, methoxy-and keto-MAs (Fig. 4A). To examine the potential involvement of HadD Mtb protein in their biosynthesis, the MAs were extracted from Mtb ∆hadD Mtb strain and analyzed. The MA content, expressed as the ratio 'MA dry weight/delipidated bacterial residue dry weight' , was similar between the wt (10.2 ± 1.1%) and the mutant (10.4 ± 1.7%) strains. Yet, the MA distribution was changed in the deletion mutant, as shown by HPTLC analysis of the MA methyl esters (MAMEs) (Fig. 4B,C). The abundance of keto-MAs was reduced by 63%. This was compensated by an increase in the α-MA content. The recovery of the wt profile upon complementation with a www.nature.com/scientificreports www.nature.com/scientificreports/ wt hadD Mtb copy suggested that this gene is directly involved in keto-MA biosynthesis 26 . The overexpression of hadD Mtb in Mtb wt strain supported this conclusion, since it induced a strong increase (of 87%) in the keto-MA relative content (Fig. 4C). This had repercussions on the relative abundance of the biosynthetically affiliated methoxy-MAs, which increased slightly (Fig. 4C).
These results altogether clearly demonstrate that HadD Mtb plays an important role in the biosynthesis of the keto-MAs. It is noteworthy that no major change in the profile of other lipids from Mtb was detected ( Supplementary Fig. S1) suggesting that hadD Mtb is not involved in another lipid biosynthesis pathways. Therefore, the significant alteration of MA distribution in Mtb ∆hadD is most likely responsible for its loss of virulence (Fig. 3A).

production of full-size keto-mycolic acids requires an active HadD Mtb protein.
To investigate further the function of HadD Mtb , the fine structures of MAs were analyzed using MALDI-TOF mass spectrometry (MS) and 1 H-NMR spectroscopy. The mass spectrum of the total MA mixture from Mtb ∆hadD displayed an increase of the relative intensities of α-MAMEs signals with respect to the wt strain (Fig. 5). More importantly, the size distribution of the keto-MAs changed significantly. The signal intensities of the long chain molecules (C 82 -C 88 ) markedly decreased, while those of the short-chain keto-MAs (C 78 , C 80 ) raised (Fig. 5). The same observations were made on the spectrum of keto-MAs purified from the deletion mutant ( Supplementary Fig. S2). The wt profiles were partially recovered in the complemented strain Mtb ΔhadD::hadD Mtb (Fig. 5, Supplementary  Fig. S2), confirming the role played by hadD Mtb deletion in this phenotype. It is noteworthy that a reduction in the proportion of the longest molecules was also detected for the methoxy-MAs in Mtb ∆hadD (Fig. 5). Moreover, when hadD Mtb was overexpressed in Mtb wt, the relative content of the long-chain keto-MAs (C 82 -C 86 ) strongly rised (Fig. 5).
Consistent with HPTLC analyses (Fig. 4B,C), these data altogether confirm the involvement of HadD Mtb in the keto-MA pathway. Furthermore, they show that HadD Mtb is implicated in the late elongation cycles during the biosynthesis of the longest chain keto-meromycolic chains. This is in agreement with the lack of HadD ortholog in other genera of the Corynebacteriales order, such as Nocardia, Rhodococcus, and Gordonia, which produce  Table S2). This suggested that the functions of HadD Mtb and HadD Msm proteins are similar. Indeed, the present work and previous data 16 show that they are both involved in dehydratation steps during late FAS-II elongation cycles. Yet, unlike hadD Mtb , the hadD Msm gene was not able to restore the virulence level of Mtb wt in Mtb ΔhadD strain (Fig. 3A,B). Furthermore, the reverse cross-complementation, i.e. the expression of hadD Mtb in M. smegmatis ∆hadD, was far to be as successful in terms of MA profile and physiology. The MA distribution was much closer to that of M. smegmatis ∆hadD than that of M. smegmatis ∆hadD::hadD Msm (Supplementary Fig. S3A,B). The same conclusions were drawn from the colony morphology observation (Supplementary Fig. S3C) as well as from the sensitivity assays to rifampicin ( Supplementary Fig. S3D), towards which M. smegmatis ∆hadD exhibits a hypersensitivity 16 . These data show that the functions of HadD Mtb and HadD Msm are not completely superimposable. This is in perfect agreement with the specificities of HadD Msm for the αand epoxy-MA biosynthesis 16 and of HadD Mtb for the keto-MA pathway. The inactivation of hadD Mtb did not induce a reduction of α-MA content in Mtb as in M. smegmatis 16 . Therefore, although HadD Mtb and HadD Msm possess similar functions, they are not strict orthologs, and the substrate specificity of HadD Mtb seems important for its role in virulence.

Accumulation of dehydratation substrates triggered by hadD Mtb inactivation. It is noteworthy
that Mtb ∆hadD accumulated a polar compound 'Z' , undetectable in Mtb wt and hardly visible in Mtb ∆had-D::hadD Mtb by HPTLC (Fig. 4B). This compound, appearing as a double band, likely corresponded to a heterogeneous mixture of molecules. After purification, the MALDI-TOF mass spectrum of compound Z displayed peak envelopes within a mass range from 1148 to 1362 Da ( Supplementary Fig. S4A) reminiscent of the classical MAME mixture (Fig. 5), but with an addition of 16 mass units. Consistent with this, the mass increment of 84 Da (corresponding to two acetyl groups) observed after per-O-acetylation of compound Z indicated the presence of two hydroxyl groups in the intact compound ( Supplementary Fig. S4A), which explained its low R f in HPTLC (Fig. 4B). This was confirmed by 1 H-NMR spectroscopy of intact compound Z generating two signals at 3.85 ppm and 3.98 ppm assigned to methines bearing both hydroxyl groups (-CHOH-). After peracetylation of compound Z, these signals shifted at 4.92 ppm and 5.13 ppm, corresponding to the chemical shifts of methines bearing O-acetyl groups (Supplementary Fig. S4B). The spin system in the 1 H-1 H 2D NMR COSY spectrum of per-O-acetylated compound Z revealed that, in the intact compound, one hydroxyl group is carried by the C-3 like in the classical MAs, whereas the additional hydroxyl group is located on the C-5 ( Supplementary  Fig. S4C,D). These data altogether clearly showed that compound Z is a mixture of 5-hydroxylated MAs. These molecules are mostly cis-cyclopropanated like regular Mtb MAs (Supplementary Table S2).
The additional hydroxyl group on the C-5 in compound Z was located on the C-3 in the precursor meromycolic chains, before the mycolic condensation step (Supplementary Fig. S5). The 3-hydroxylated meromycolic acids were taken over by the mycolic condensation system leading to the synthesis of abnormal 5-hydroxylated MAs. The accumulation of 3-hydroxylated intermediates corresponding to dehydratase substrates strongly suggests that, in the absence of HadD Mtb protein, the dehydratation step of the late FAS-II elongation cycles is partially blocked, preventing the formation of mature meromycolic chains (Supplementary Fig. S5). The 3-hydroxylated www.nature.com/scientificreports www.nature.com/scientificreports/ α-meromycolic chains might come from the group of diethylenic precursors of oxygenated MAs (Fig. 6), which, accumulating abnormally, would later be cyclopropanated by default like the regular α-meromycolic chains.
In conclusion, HadD Mtb most likely catalyzes the 3-hydroxyacyl-ACP dehydratation step of the late elongation cycles during the biosynthesis of the oxygenated meromycolic chains.

Discussion
The present study reports the existence in Mtb of a putative ortholog, which we named HadD Mtb , of the recently discovered (3R)-hydroxyacyl-ACP dehydratase HadD Msm of the FAS-II system from M. smegmatis 16 . The in-depth analysis of the distribution and fine structure of the MAs produced by Mtb hadD deletion mutant revealed that HadD Mtb is involved in the oxygenated MA biosynthesis, and more particularly dedicated to the keto-MA pathway. Indeed, in the mutant, the overall keto-MA content was strongly reduced. The fact that the relative abundance of the longest keto-MAs decreased in Mtb ∆hadD Mtb whereas it increased in Mtb wt::hadD Mtb overexpression strain shows that HadD Mtb is involved in late FAS-II elongation cycles during the biosynthesis of the keto-meromycolic chains. www.nature.com/scientificreports www.nature.com/scientificreports/ This conclusion is supported by the lack of HadD ortholog in the genera of the Corynebacteriales order that produce only intermediate size and not full size mycolic acids as found in mycobacteria 16 . The abnormal 5-hydroxylated MAs observed in HadD Mtb deficient strain result from the condensation of 3-hydroxy-meromycoloyl chains with a carboxyacyl chain (Supplementary Fig. S5). Their accumulation together with the belonging of HadD Mtb to the hydratase 2 protein family and its high sequence similarity with HadD Msm leads to the conclusion that HadD Mtb most likely catalyzes the 3-hydroxyacyl-ACP dehydratation step of these FAS-II elongation cycles (Fig. 6). We had previously discovered two (3 R)-hydroxyacyl-ACP dehydratases of the FAS-II system from M. tuberculosis, HadAB and HadBC 13 . Thus, HadD Mtb constitutes a third dehydratase of this system. It has been shown that HadAB is involved in the early elongation cycles common to both α and oxygenated MA pathways, while HadBC is required for the late cycles leading to the biosynthesis of the sole oxygenated MAs, which are 4-6 carbon longer than α-MAs 13,27 . Since the keto-and methoxy-MAs are biosynthetically affiliated 28 , the MA distribution in Mtb ∆hadC strain was interpreted in terms of α/oxygenated ratio 27 . Yet, it is noteworthy that, in this mutant, the keto-MA content remains stable whereas that of the methoxy-MAs drops dramatically 27 . In the light of our new findings, this strongly suggests that HadBC is preferentially dedicated to the methoxy-MA biosynthesis and HadD Mtb to the keto-MA pathway. However, the functions of both enzymes are likely partially redundant since their individual inactivation does not totally inhibit the production of methoxy-or keto-MAs. This is supported by the variations in methoxy-MA content and size distribution observed in Mtb wt::hadD Mtb and in Mtb ∆hadD Mtb , respectively (Figs. 4 and 5), and in keto-MA fine structure in Mtb ΔhadC strain 27 , which may also be partly due to a tight regulation between both biosynthesis pathways. These surveys allow us to draw a biosynthesis scheme for the three MA classes from Mtb, which details the specific roles of the three FAS-II dehydratases (Fig. 6).
The functions of HadD Mtb and HadD Msm in M. smegmatis appear closely related since they both catalyze the dehydratation step during late elongation cycles 16 . Yet, the protein from Mtb has a substrate specificity distinct from that of M. smegmatis, where it is required for the biosynthesis of αand epoxy-MAs 16 . The phenotypic analyses of the cross-complemented M. smegmatis ∆hadD mutant in terms of MA profile, colony morphology and rifampicin sensitivity, and of the cross-complemented Mtb ∆hadD mutant in terms of virulence level confirmed this conclusion and definitely showed that hadD Mtb and hadD Msm are not strict orthologs, despite the partial genomic conservation of their chromosomal regions. This suggests that there has been a functional divergence of both genes after the speciation event. Indeed, during its adaption to a distinct environment, the ortholog in a new species may undergo neofunctionalization, resulting in a species-specific function for this gene 29 .
We showed that Rv0504c gene encoding HadD Mtb is not essential for the survival of the tubercle bacillus in axenic culture. However, we observed that it has an important impact on the fitness of bacteria, their organization into biofilms and colonies, as well as their tolerance to low temperature (30 °C). This is certainly linked to the determinant role of MAs in the architecture and the fluidity of the envelope due to their strategic location within the mycomembrane 2,30 . In agreement with our data, the growth of an Mtb strain overproducting mmaA3 MA-methyltransferase gene and lacking keto-MAs was impaired at reduced temperature (32 °C) 31 . MAs are also known to be important for the formation of mycobacterial biofilms, where the extracellular matrix contains large amounts of free MAs 32 . In particular, the production of keto-MAs in which HadD Mtb is involved is essential for biofilm growth 33 . MAs represent also key Mtb pathogenicity factors, their fine structures strongly potentiating the immune response to infection 2 . Here, we show that HadD Mtb function greatly influences the virulence level of Mtb in the mouse model of infection. This is consistent with previous findings showing that a Mtb hma (mmaA4) mutant devoid of oxygenated MAs is attenuated in mice 34 and that keto-MAs are critical for Mtb growth within the natural host cells 31 . Additionally, oxygenated MAs play a role in the selective repression of macrophage IL-12p40 cytokine production aimed to evade elimination by the host immune system 35 , and the hma gene required for their biosynthesis is actively expressed during human pulmonary tuberculosis 36 . Importantly, the recently approved drug delamanid, active against MDR-TB, kills Mtb by blocking the oxygenated MA production 37 . Altogether, these data indicate that targeting this metabolism would constitute a relevant strategy for the development of new therapeutics against Mtb. Because of its involvement in the keto-MA biosynthesis pathway, its role in Mtb virulence and its specificity to mycobacterial cells 16 , HadD protein represents a promising pharmaceutical target. Besides, targeting non-essential enzymes of the tubercle bacillus should limit the occurrence of antibiotic resistance mechanisms. (hadD Mtb ) gene in Mtb H37Rv (ATCC 27294) strain was done using a previously described method 19 . Briefly, the Rv0504c deletion delivery vector was constructed by amplifying the upstream (0.95 kb) and downstream (1.1 kb) regions of the gene using the primer pairs EUF24 (5′-GCTGCAGCTTCTTCGACGTGGACAACA-3′) and EUR24 (5′-CAAGCTTGCCGATCAGTGTCTGGGCTTCT-3′) and EUF25 (5′-CAAGCTTGCCGAGATC CGAAGCGAAGTTA-3′) and EUR25 (5′-CGGTACCTTTCGGCTGACCCTTATTG-3′), respectively. (2020) 10:2112 | https://doi.org/10.1038/s41598-020-58967-8 www.nature.com/scientificreports www.nature.com/scientificreports/ PCR-amplified fragments were cloned into p2NIL 19 using the restriction sites in the primers (PstI, HindIII and Kpn I, underlined in the above sequences) and the sequence was verified. The PacI cassette containing lacZ, sacB, and hyg from pGOAL19 19 was introduced to construct the final vector pTACK0504G. The plasmid was electroporated into Mtb and single crossovers were isolated. Double crossovers were isolated from the single crossover strain as previously described 19 . Colonies were screened for the presence of the wild-type (wt) or deletion alleles by PCR using primers 0504D1 (5′-AGCCTCTAGACGCCAATCAC-3′) and 0504D2 (5′-GGCTCAAGGTTCAGCTTGTC-3′). The deletion was checked by PCR and sequencing. For complementation of Mtb ∆hadD strain, the wt copy of Rv0504c gene, containing its natural promoter (186 bp before the start codon), was amplified with the primer pairs 0504C1 (5′-AGCCTCTAGACGCCAATCAC-3′) and 0504C2 (5′-CAAGCTTCGTCATTGAACGGACCCTAC-3′) that incorporates a HindIII restriction site and cloned into pSC-A (Agilent). The HindIII fragment from pUC-GM-Int containing the mycobacteriophage L5 integrase, att site, and Gm resistance 40 was inserted to make the complementing vector. The resulting plasmid pUC-hadD Mtb and the empty plasmid pUC (as a control) were used to transform both Mtb wt and Mtb ∆hadD strains (Supplementary Table S3). For cross-complementation experiments, pUC-hadD Msm plasmid carrying a wt copy of MSMEG_0948 (hadD Msm ) gene 16 was used to transform Mtb ∆hadD and pUC-hadD Mtb plasmid was used to transform M. smegmatis ΔhadD strain previously constructed 16 (Supplementary Table S3). Data were compared to those obtained for M. smegmatis mc 2 155 (wt) strain as well as M. smegmatis ΔhadD complemented by pUC-hadD Msm and previously described 16 . culture conditions and phenotyping assays. For biofilm growth and lipid analyses, Mtb strains were cultured as pellicles in glass bottles for 4-5 weeks at 37 °C in 7H9 broth (Difco) containing 0.2% glycerol, 10% Middlebrook ADC (Difco) and 10 µg/ml gentamycin. Cultures were inoculated with identical volumes of precultures in exponential growth phase grown in the same medium then adjusted at OD 600 ~ 1. For the growth curves, planktonic cultures were realized under shaking (120 rpm) in the above medium supplemented with 0.05% (w/v) Tween-80, and OD 600 was measured at different time points. For colony morphology and susceptibility to temperature assays on Mtb strains, liquid precultures were done in the latter medium (with Tween-80) then adjusted to the same OD and serially diluted. Five µl aliquots of each dilution were spotted on Middlebrook 7H11 medium (Sigma) supplemented with 0.5% glycerol, 10% Middlebrook OADC (Difco) and 10 µg/ml gentamycin. Cultures were incubated for at least 3 weeks at 37 °C or at 30 °C (for temperature testing). The minimum inhibitory concentrations for several antibiotics and SDS were determined in 7H9 broth supplemented with glycerol by using a colorimetric microassay based on the reduction of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT, Sigma-Aldrich) into formazan by metabolically active cells, as described 41 . Colony morphology and rifampicin susceptibility assays on M. smegmatis strains were performed as previously described 16 . Lipid extractions, (HP)TLC analyses and purification. The MAs and the total extractable lipids were extracted and analyzed as described previously 16 , except that the (HP)TLC plates for MAME separation were developed in ether/diethyl ether (9:1, v/v, five runs). Purification of MAMEs and compound Z was performed by preparative TLC using silica gel 60 plates (Merck) developed in the above eluent; the compounds were then scraped off and extracted from silica gel three times with diethyl ether.

Methods
Lipid structural analyses. Peracetylation of compound Z was performed in pyridine:acetic anhydride 1:1 for 1 h at 100 °C. After drying, three extractions with H 2 O:diethyl ether (1:1) were done; the ether phases were collected, washed three times with water and dried. MALDI-TOF MS analyses were performed in the positive ionization and reflectron mode, using the 5800 MALDI-TOF/TOF Analyzer (Applied Biosystems/ABsciex) equipped with a Nd:YAG laser (349 nm wavelength) as described previously 16 . 1 H-NMR and 1 H-1 H-NMR COSY spectra were recorded in CDCl 3 at 298° K using a 600-MHz Bruker Avance III spectrometer (Bruker Biospin) equipped with a TCI cryoprobe. Chemical shift values were referenced to CHCl 3 resonance (δH 7.26 ppm). The quantification of MAME unsaturations was performed with TopSpin 3.5 pl7 software.

Mouse infection experiments.
Virulence studies were performed in SCID mice. Briefly, groups of 6 week old C.B-17/Icr SCID mice (Charles River) were intravenously infected with a bacterial suspension containing 1-2 × 10 5 CFU/mouse. One day and 28 days after infection, mice were euthanized, and spleen and lungs were homogenized in 2 ml tubes containing 500 µl Sauton medium and 2.5 mm diameter glass beads using an MM300 apparatus (Qiagen). CFU numbers in target organs were determined by plating 5-or 10-fold serial dilutions of organ homogenates on solid medium and incubation at 37 °C. These studies were approved by the Institut Pasteur Safety Committee (Protocol 11.245; experimentation authorization number 75-1469), in accordance with European and French guidelines (Directive 86/609/CEE and Decree 87-848 of 19 October 1987), and implicating approval from local ethical committees (CETEA 2013-0036 and CETEA dab180023).

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary  Information files).